mirror of
https://github.com/LostRuins/koboldcpp.git
synced 2026-08-26 06:31:13 +02:00
Wagner Bruna
b9b3bbcec1
* refactor: add new module for backend-specific code * replace ifdefs on rwkv_v3 * replace ifdefs on llama_v2 * replace ifdefs on llama_v3 * replace ifdefs on gpt2_v3 * replace ifdefs on gptj_v3 * replace ifdefs on mpt_v3 * replace ifdefs on neox_v3 * replace ifdefs on whisper * adjust build for kcpp_backend Mostly a: sed 's/ gpttype_adapter\([^ ]*\)\.o / gpttype_adapter_default.o kcpp_backend\1.o /g' Makefile and include variant objects for kcpp_backend.
42 lines
1.5 KiB
C
42 lines
1.5 KiB
C
#pragma once
|
|
|
|
#include "ggml-backend.h"
|
|
#include "ggml_v2.h"
|
|
#include "ggml_v3.h"
|
|
|
|
// backends with GGML_USE_CUDA
|
|
#define KCPP_BACKENDS_USE_CUDA "cuda|rocm"
|
|
|
|
// backends that support tensor split
|
|
#define KCPP_BACKENDS_TENSOR_SPLIT "cuda|rocm|vulkan"
|
|
|
|
// backends that support blas
|
|
#define KCPP_BACKENDS_BLAS "blas|cuda|rocm|vulkan|sycl"
|
|
|
|
// checks if the provided backend (or the first one) matches a |-separated name list
|
|
int kcpp_backend_check(const char* name_list, ggml_backend_t backend = nullptr);
|
|
|
|
ggml_backend_dev_t kcpp_backend_get_device(int index);
|
|
|
|
// per-backend aux functions
|
|
|
|
void kcpp_backend_cuda_ggmlv2_transform_tensor(ggml_v2_tensor * tensor);
|
|
|
|
void kcpp_backend_cuda_ggmlv3_assign_buffers(struct ggml_v3_tensor * tensor);
|
|
void kcpp_backend_cuda_ggmlv3_assign_buffers_force_inplace(struct ggml_v3_tensor * tensor);
|
|
void kcpp_backend_cuda_ggmlv3_assign_buffers_no_scratch(struct ggml_v3_tensor * tensor);
|
|
void kcpp_backend_cuda_ggmlv3_free_data(struct ggml_v3_tensor * tensor);
|
|
void kcpp_backend_cuda_ggmlv3_free_scratch(void);
|
|
void kcpp_backend_cuda_ggmlv3_set_main_device(int device);
|
|
void kcpp_backend_cuda_ggmlv3_set_mul_mat_q(bool mul_mat_q);
|
|
void kcpp_backend_cuda_ggmlv3_set_scratch_size(size_t scratch_size);
|
|
void kcpp_backend_cuda_ggmlv3_set_tensor_split(const float * tensor_split);
|
|
void kcpp_backend_cuda_ggmlv3_transform_tensor(void * data, struct ggml_v3_tensor * tensor);
|
|
|
|
void kcpp_backend_cuda_set_mul_mat_q(int use_mmq);
|
|
|
|
void kcpp_backend_hip_initialize();
|
|
|
|
bool kcpp_backend_metal_supports_family(ggml_backend_t backend, int family);
|
|
|